Rajat M R
Community@atomicrajat · Bangalore
Solutions Architect - Edge AI | Deepstream | Accelerated AI | Deep Learning | TinyML | NVIDIA Jetson
Agent Skills by Rajat M R
Showing 28 vetted skills indexed across 1 GitHub repositories.
atomicrajatjetson-customize-clocks
Configure Jetson CPU, GPU, and EMC clock ceilings by editing BPMP DTB and nvpower.sh pre-flash.
atomicrajatvss-deploy-profile
Deploy, verify, debug, and tear down NVIDIA VSS compose profiles on GPU hosts.
atomicrajatdeepstream-profile-pipeline
Profile DeepStream pipelines with Nsight Systems and derive performance configs from measurements.
atomicrajatdeepstream-dev
Build NVIDIA DeepStream video analytics pipelines with the Python pyservicemaker API.
atomicrajatjetson-inference-mem-tune
Recommends inference runtimes and memory launch flags for LLM/VLM serving on NVIDIA Jetson devices.
atomicrajatvss-deploy-detection-tracking-2d
Deploy, debug, and operate the RTVI-CV 2D detection and tracking microservice via Docker and REST API.
atomicrajatjetson-customize-nvpmodel
Edits nvpmodel power-mode configurations on Jetson Orin and Thor BSP overlays.
atomicrajatvss-ask-video
Ask visual questions about recorded video clips through the VSS agent's video_understanding tool.
atomicrajatjetson-print-device-info
Prints Jetson device model, L4T version, kernel, OS version, and power mode.
atomicrajatvss-query-analytics
Query video-analytics incidents, alerts, and sensor metrics through the VA-MCP server.
atomicrajatnvidia-skill-finder
Discover and recommend NVIDIA agent skills from the live catalog for NVIDIA-related tasks.
atomicrajatdeepstream-run-mv3dt
Run and operate the DeepStream Multi-View 3D Tracking reference app with samples, custom datasets, and Kafka metadata.
atomicrajatvss-summarize-video
Summarize recorded video clips via the LVS microservice with a VLM fallback.
atomicrajatjetson-diagnostic
Captures read-only Jetson health snapshots covering identity, memory, GPU, thermal, power, and services.
atomicrajatvss-setup-behavior-analytics
Deploys the vss-behavior-analytics container standalone with chosen entrypoint, config, and calibration.
atomicrajatrtvi-vlm-customize-model
Reconfigures the VLM endpoint across all three VLM consumers in the NVIDIA VSS Alerts Blueprint.
atomicrajatvss-manage-alerts
Manage VSS alert subscriptions, real-time monitoring, Slack notifications, and incident queries.
atomicrajatdeepstream-generate-pipeline
Generate validated gst-launch-1.0 pipelines for NVIDIA DeepStream video inference and tracking.
atomicrajatvss-manage-video-io-storage
Manage VIOS REST API operations for sensors, streams, clips, snapshots, and video storage.
atomicrajatvss-deploy-dense-captioning
Deploy the standalone RT-VLM dense-captioning microservice and exercise its REST API endpoints.
atomicrajatvss-setup-video-analytics-api
Deploy the vss-video-analytics-api REST service standalone with Elasticsearch and optional Kafka.
atomicrajatvss-generate-video-report
Generates structured video analysis and incident-range reports from VSS clips and analytics.
atomicrajatdeepstream-sop
Builds and debugs the DeepStream SOP inference microservice for VLM-based procedure compliance monitoring.
atomicrajatjetson-optimize-memory
Reclaim DRAM on Jetson devices by disabling unused display, camera, and SWIOTLB carveouts.
Frequently Asked Questions About Rajat M R
FAQPage SchemaWhat tasks can I accomplish with atomicrajat's NVIDIA skill registry?▼
You can tune Jetson clocks, power modes, and memory carveouts; build, profile, and benchmark DeepStream video pipelines; run Multi-View 3D Tracking; and deploy the full VSS blueprint for video search, summarization, alerts, dense captioning, and VLM-based visual question answering.
Who is the target audience for these skills?▼
Edge AI solutions architects, embedded engineers working on NVIDIA Jetson Orin/Thor targets, and video analytics developers building DeepStream GStreamer pipelines or operating NVIDIA VSS warehouse, smart-city, and factory SOP-compliance deployments.
What are the runtime prerequisites for DeepStream profiling?▼
DeepStream SDK 9.0 on Ubuntu 22.04 or 24.04 inside the nvcr.io/nvidia/deepstream:9.0-triton-multiarch dev container, with Nsight Systems 2024+ (nsys) and nvidia-smi on PATH. It runs fully headless; the samples-multiarch image must be avoided for profiling.
Are these skills open source and what licenses apply?▼
Yes. Most skills ship under Apache-2.0, while NVIDIA-authored DeepStream and skill-finder entries use dual CC-BY-4.0 AND Apache-2.0 licensing. VSS blueprint skills reference the public NVIDIA-AI-Blueprints video-search-and-summarization GitHub repository.
How do I optimize LLM or VLM inference memory on Jetson?▼
Use jetson-inference-mem-tune to select a serving stack (vLLM, SGLang, llama.cpp, or TensorRT Edge-LLM) with per-runtime memory flags, and jetson-optimize-memory to reclaim DRAM by disabling unused subsystems across MB1/MB2 BCT, kernel reserved-memory, and SWIOTLB.